Better at What the Thing Is Once It's In

The most useful thing I did this month was spend two days trying to convince myself to stop.

The question, put as bluntly as I could stand it: mymind is already this good. Why would anyone use mine?

Not a positioning exercise. A genuine attempt to find out whether I should be building this at all. The rule I set was that I was not allowed to protect the hypothesis, and that a finding in the competitor's favour counted for exactly as much as one in mine.

Where mymind is straightforwardly better

Let me state this without hedging, because a competitive analysis that finds the competitor worse at everything is marketing with footnotes.

Getting a thing in. mymind has a native iOS application with a share extension. From a screenshot to saved is two taps, from anywhere in the system. Mine is a web application, and iOS has no web share target. This is not something I can out-design. It is a platform capability that does not exist for web applications, and only a native share extension can provide it.

Understanding an image. mymind runs OCR, including handwriting. Mine does not. A screenshot in my product is findable by its name, its date and its kind, but not by the words printed inside it. That is the single largest retrieval gap in my product. I have left it open deliberately, and open is open.

Automatic organisation. AI tags, colour extraction, and around 33 automatically maintained Stacks. Real work, done well, requiring nothing from the user.

That is a serious product, actively developed, run by people with taste. Anyone choosing between the two on capture alone should choose theirs.

Where the position becomes checkable

Then I read their help centre rather than their marketing, and a different picture appeared. One they are entirely upfront about, which is to their credit.

On whether entire articles are indexed: "We currently do not index the entire text of websites saved to your mind." They extract key phrases, names, places, core concepts, and deliberately drop the rest to keep search fast.

On PDFs, more explicitly: "The entire contents of a PDF are not currently indexed." Tags, title and an AI summary are searchable. The text inside the document is not. Their stated reason is sensible: full indexing would produce thousands of keywords per document and slow everything down.

On getting your material out: there is no import feature, by policy. Export exists but runs from desktop Chrome or Edge, and cannot be re-imported after an account is closed.

Table 1. Sources are mymind's own help centre, checked September 2026. Blank rows are genuinely unknown rather than assumed.

The position, stated as narrowly as it deserves

mymind is better at getting a thing in. Mine is better at what the thing is once it is in, and at getting it back out.

That is the whole claim. It is narrow. It is checkable. It does not say I am winning, and it would be false if I said one word more.

What that means concretely

Typing 용역 finds a contract whose filename is 계약서 최종본.pdf and whose only occurrence of that word is inside the file's bytes. The search reads the document's own text, the page's title and description, an archive's contents, the host of a link. Every word you type must literally appear, so when it finds nothing, that is a fact rather than an opinion.

And the way out is handed back per kind: a line's words, a link's address, a picture's original pixels rather than the card-sized preview, and machine-read text always labelled as machine-read.

THE SAME CONTRACT, TWO SYSTEMS

  file:  계약서 최종본.pdf
  the word "용역" appears ONLY inside the bytes,
  on page 4, and nowhere in the filename or title

  key-phrase index    →  not found
     (the phrase was not judged important enough to keep)

  literal full-text   →  found
     match shown in context, page located

THE SAME CONTRACT, TWO SYSTEMS

  file:  계약서 최종본.pdf
  the word "용역" appears ONLY inside the bytes,
  on page 4, and nowhere in the filename or title

  key-phrase index    →  not found
     (the phrase was not judged important enough to keep)

  literal full-text   →  found
     match shown in context, page located

THE SAME CONTRACT, TWO SYSTEMS

  file:  계약서 최종본.pdf
  the word "용역" appears ONLY inside the bytes,
  on page 4, and nowhere in the filename or title

  key-phrase index    →  not found
     (the phrase was not judged important enough to keep)

  literal full-text   →  found
     match shown in context, page located

This is not a cleverness gap. It is a direct consequence of two different architectural bets about what to store, and either bet is reasonable right up until the day you need the other one.

What I learned about doing this properly

Three things about the method, which I think transfer to any competitive read.

Read the help centre, not the landing page. The landing page is written to be persuasive and is carefully true. The help centre is written to reduce support tickets, which means it has to be specific, and specificity is where the real capability surface lives. Every hard fact in the table above came from a support article.

Find the trade, not the gap. mymind does not index PDF contents, and this is not an oversight. They state the reason: it would flood search with thousands of keywords per document. That is a real trade and their answer is defensible. My answer is different because my search is literal rather than ranked, so extra keywords cost nothing. Neither of us is wrong. We made different architectural bets and inherited their consequences.

Write the competitor's advantages in your own words, first. If you cannot state their strongest case better than their marketing does, you do not understand the market you are in. I spent the first of the two days doing nothing but this.

Two things I could not design around

iOS share-sheet capture is impossible for a web application. No workaround exists. On Android, file sharing would need a POST multipart target intercepted by a service worker. That is cheap now, but it changes the GET path that works today, so it is classified as unknown and deliberately not built.

Korean-only is an acquisition constraint. Anyone who lands on my product and does not read Korean is lost on the first screen. That is not a code problem. It is a decision about who this is for, and I have not made it.

Why publish a competitor's advantages on my own journal

Two reasons, and the second is the real one.

The first is practical. A position you have not tested against the strongest available alternative is not a position. It is an assumption. I now know exactly which of my claims survive contact with mymind and which do not, and the surviving set is smaller, narrower and much more defensible than the one I held two days earlier.

The second is that the alternative is worse for us. A studio that only publishes the flattering comparison produces material nobody can use, including its own future self. In two years I want to be able to hand someone this page and still stand behind every line of it. That is only possible if the unflattering half is in it.

Better at one specific thing, worse at another specific thing, and precise about which is which. That is a position. Everything above it is advertising.

Sources

• mymind. Why mymind does not have an import feature (for now). mymind.com/import-feature

• mymind help centre. Does mymind index entire websites and articles? Checked September 2026.

• mymind help centre. Does mymind index PDFs? Checked September 2026.

Related reading

• The Last Step Everybody Skips

• One Warm Colour

• Never Create a User Action So You Can Measure It

• The Door That Was Never Open

More from the Moc Mallen Studio journal